1400 Part VIII / Learning, Memory, Language and Cognition
Figure 56–5 Artificial activation of neurons that respond
preferentially to rightward motion causes a monkey to
decide that motion is rightward.In the experiment, an elec-
trode is placed in the middle of a patch of neurons in area
MT that prefer the same direction of motion, say rightward.
The random dot motion is shown in the receptive field of these
neurons. A weak alternating current is applied on half of the
trials during the presentation of the random dots movie. The
amount of current activates about 200 to 400 neurons within 50 to
100 μm of the electrode tip. On trials with microstimulation, the
monkey is more likely to choose the preferred direction of
the simulated neurons. The effect is most pronounced when the
decision is more difficult (middle red arrow). (Adapted, with
permission, from Ditterich, Mazurek, and Shadlen 2003.)
–50 2
向右选择的百分比
–25 0
运动强度(抑制性百分比)
使用微刺激
无微刺激
更强的
向左
运动
更强的
向右
运动
neurons were likely to share the same receptive field
and the same direction preference. Newsome had the
monkey decide between this direction and its oppo-
site. For example, if these neurons preferred rightward
motion, the weak currents caused the monkey to decide
more often in favor of right (Figure 56–5).
We now refer to such weak stimulation, designed
to affect a cluster of neurons within a 50- to 100μm
radius, as microstimulation. Notably, microstimulation
did not cause a hallucination of visual motion. It biased
the monkey’s decisions, which were guided mainly by
the random dot motion stimulus. The monkey did not
respond when the stimulus was not shown, and micro-
stimulation did not affect the monkey’s decisions when
the random dots were presented at a location of the
visual field outside the receptive field of the stimulated
neurons. The microstimulation exerted its largest effect
on choices when the motion strength was weakest. The
stimulated neurons simply added a small amount of
evidence for rightward motion, which is effectively
evidence against leftward motion, as discussed below.
The microstimulation experiment shows that the
direction-selective neurons in area MT contribute
evidence to the perceptual decision. However, the
stimulated neurons do not necessarily need to affect
the decision directly; they only have to participate in
a neural circuit that lies in a causal chain. In addition,
many more neurons in MT were not affected by the
electrical stimulation but nonetheless responded to
the same random dot patch in the same direction-
selective manner. They are in other columns with
receptive fields that are not centered on the stimulus
but overlap it. If the electrode is moved to stimulate
these neurons, they too cause the monkey to choose the
preferred direction more often. These findings imply
that in any one experiment the microstimulation only
affects a small fraction of the neurons that contribute
to the decision. Most respond at their usual firing rates
to the random dot motion. The microstimulation only
changes the total signal that the brain uses to make its
decision by a small amount. No wonder the effect is
only evident when the decision is difficult.
There is an important principle to be learned here.
Had Newsome used only the easier conditions, the elec-
trical stimulation would have yielded a null effect, and
thus, the causal relationship between the neural activ-
ity and behavior would not have been established. The
same pattern of effects has recently been established
using techniques to turn neurons off. Silencing induces
a bias in choices against the direction of the silenced
neurons, but this too is only apparent on trials when
the motion is difficult. Without evidence for sufficiency
or necessity, a neuroscientist might conclude that the
neurons in MT do not cause changes in perceptual deci-
sions. This would be a mistake, notably one that is likely
to be made in any experiment in which perturbations
are restricted to a subset of the neurons involved in a
computation. That is the rule, not the exception, for
studies of higher cortical functions. It is only mitigated
by studying behavior in conditions when a small differ-
ence to the total pool of neural signals might make a dif-
ference, as in the difficult (low signal-to-noise) regime
employed in Newsome’s experiments.
To summarize so far, the perceptual decision arises
from a simple decision rule: the application of a criterion to
the noisy evidence supplied by noisy direction-selective
neurons in the visual cortex. We have characterized the
noisy evidence as a single number: the difference in the
mean firing rates from two opposing pools of direction-
selective neurons. This account leaves out two important
points: The operations that establish the decision variable
must be carried out by neurons that receive information
directly or indirectly from area MT, and these operations
take time. As we will see, time is the key to understanding
Kandel-Ch56_1392-1416.indd 1400 19/01/21 9:33 AM